This function fit a hierarchical or a fixed-effect model, using Bayeisan sampling. We use pMCMC, with a suite of DE-MCMC, DGMC, and simply, crossover (i.e., DE-MC), mutation, or migration operators. Note that the latter two operators essentially are random-walk Metroplolis, so they will be very inefficient, if been applied alone, even with our fast C++ implementation.
run(samples, report = 100, ncore = 1, pm = 0, qm = 0, hpm = 0,
hqm = 0, gammamult = 2.38, ngroup = 5, force = FALSE,
sampler = "DE-MCMC", slice = FALSE)CheckConverged(samples)
a sample list generated by calling DMC's samples.dmc.
how many iterations to return a report
parallel core for run_many
probability of migration
probability of mutation
probability of migration at the hyper level
probability of mutation at the hyper level
a tuning parameter, affecting the size of jump
number of distributed groups
set force to FALSE for turning off recalculation of PDA. Set it as an integer between 1 and 10, forcing to re-calculate new likelihood, every e.g., 1, 2, 3 step.
which sampler to run MCMC, "DE-MCMC" or "DGMC"
use for debugging blocked sampling